Kaldi requires metadata to run advanced algorithms. The metadata consists of several text file lists, such as the list of train/test/dev utterances (sorted key utterance pairs), speaker to utterance file spk2utt (speaker utterances lists) and utterance to speaker file utt2spk (utterance speakers lists). The idea is to add this metadata to prospective databases that could be used with Kaldi, such as AMI, NIST and so far.